SYNTOLOGY HomeExplorerAtlasCodeMethodologyAboutDevelopersFeedPricing
Paper · aaai_16558 · AAAI · 2021

DocParser: Hierarchical Document Structure Parsing from Renderings

Ce Zhang, Stefan Feuerriegel, Johannes Rausch, Octavio Martinez, Fabian Bissig

Open in the Atlas

Code that ran

14 of 17 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 0 honoured the contract we drafted, 1 violated it and 13 have no contract verdict; 0 failures of our instrument; 3 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.

"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/aaai_16558.json · badge: /api/ran/aaai_16558.svg

Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.

RepositoryRoleRan
DS3Lab/DocParser canonical (classifier label)
classifier basis: census
14 ran · 3 no recorded run
FunctionStatusWhere it lives
adjust_and_move_row_and_column_borders Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/postprocess_table_structure.py
code served (permissive licence) · get_code("a5e0e7250263a2a8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

ARGS = {
    "annotations": [
        {
            "category": "table_row",
            "row_nr": 0,
            "bbox": [10.0, 20.0, 100.0, 15.0]
        },
        {
            "category": "table_row",
            "row_nr": 1,
            "bbox": [10.0, 50.0, 100.0, 15.0]
        },
        {
            "category": "table_col",
            "col_nr": 0,
            "bbox": [10.0, 20.0, 40.0, 50.0]
        },
        {
            "category": "table_col",
            "col_nr": 1,
            "bbox": [55.0, 20.0, 40.0, 50.0]
        },
        {
            "category": "table_col",
            "col_nr": 2,
            "bbox": [100.0, 20.0, 40.0, 50.0]
        }
    ]
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
convertToAbsoluteValues Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/objdetmetrics_lib/utils.py
code served (permissive licence) · get_code("5fa22628f1bde0ae")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyzing the function:
# - size: tuple/list of 2 elements (width, height of image)
# - box: tuple/list of 4 elements (center_x, center_y, width, height in normalized coords)
# Both are accessed with indexing and converted to float

# Create valid arguments
size = (640, 480)  # Image dimensions: width=640, height=480

# box format appears to be normalized coordinates:
# [center_x, center_y, box_width, box_height] all in range [0, 1]
box = [0.5, 0.5, 0.3, 0.4]  # Center at (0.5, 0.5), width=0.3, height=0.4

ARGS = {
    "size": size,
    "box": box
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family int — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
convertToRelativeValues Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/objdetmetrics_lib/utils.py
code served (permissive licence) · get_code("478a1014eb6ca877")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned tuple, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyzing the function:
# - size: indexed as size[0] and size[1], should be a tuple/list with 2 elements
# - box: indexed as box[0], box[1], box[2], box[3], should be a tuple/list with 4 elements
#   representing bounding box coordinates (x_min, x_max, y_min, y_max)

# Create valid arguments
size = (640, 480)  # width, height of image/canvas
box = (10.0, 100.0, 20.0, 150.0)  # (x_min, x_max, y_min, y_max) - bounding box coordinates

ARGS = {
    "size": size,
    "box": box
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type tuple, ndim 1, dtype family float — a tuple output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_all_children_ids_with_child_dictionary Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/data_utils.py
code served (permissive licence) · get_code("f523e10c5c818bf8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

# The function recursively traverses a tree structure represented as a dictionary
# where keys are parent IDs and values are lists of child IDs.
# It returns all descendant IDs for a given parent.

# Create a simple tree structure:
#       1
#      / \
#     2   3
#    /
#   4

ann_children_ids = {
    1: [2, 3],
    2: [4],
    3: [],
    4: []
}

parent_id = 1

ARGS = {
    "parent_id": parent_id,
    "ann_children_ids": ann_children_ids
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_all_children_ids_with_child_dictionary Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/data_utils.py
code served (permissive licence) · get_code("d3c0e6c2dc9389fe")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

ARGS = {
    'parent_id': 1,
    'ann_children_ids': {
        1: [2, 3],
        2: [4, 5],
        3: [6],
        4: [],
        5: [],
        6: []
    }
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_col_overlaps Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/postprocess_table_structure.py
code served (permissive licence) · get_code("4a8657f9c3c5f19b")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Construct valid arguments for get_col_overlaps
# The function expects:
# - col_anns: dict mapping column numbers to annotation dicts with 'bbox' key
# - cell_ann: dict with 'bbox' key
# where bbox is in COCO format [x0, y0, w, h]

col_anns = {
    0: {'bbox': [10, 20, 50, 30]},
    1: {'bbox': [60, 20, 40, 30]},
    2: {'bbox': [110, 20, 35, 30]},
}

cell_ann = {'bbox': [30, 20, 60, 30]}

ARGS = {
    'col_anns': col_anns,
    'cell_ann': cell_ann,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_matches_per_image Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/eval_utils.py
code served (permissive licence) · get_code("5f84a5f22fe5f441")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned dict. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

# Set seed for reproducibility
torch.manual_seed(0)

# Build sample data structures based on function usage

# ious: list of IoU thresholds
ious = [0.5, 0.75]

# detected_relations_per_image: dict mapping image names to list of relations
# Each relation is (rel_obj_id, rel_subj_id, rel_class)
detected_relations_per_image = {
    'img_001': [
        (1, 2, 'COMES_BEFORE'),
        (3, 4, 'IS_PARENT_OF'),
    ],
    'img_002': [
        (5, 6, 'COMES_BEFORE'),
    ],
}

# relations_groundtruth_per_image: dict mapping image names to dicts with 'relations' key
# Each relation is (rel_obj_id, rel_subj_id, rel_class)
relations_groundtruth_per_image = {
    'img_001': {
        'relations': [
            (1, 2, 'COMES_BEFORE'),
            (3, 4, 'IS_PARENT_OF'),
            (7, 8, 'COMES_BEFORE'),  # false negative
        ]
    },
    'img_002': {
        'relations': [
            (5, 6, 'COMES_BEFORE'),
        ]
    },
}

# epoch_detection_results: dict with 'iou' key mapping to IoU-specific results
# Each IoU maps to a dict with 'matches_per_img' key
# matches_per_img maps image names to dicts mapping detected object IDs to GT object IDs
epoch_detection_results = {
    'iou': {
        '0.5': {
            'matches_per_img': {
                'img_001': {1: 1, 3: 3},  # detected_id -> gt_id mapping
                'img_002': {5: 5},
            }
        },
        '0.75': {
            'matches_per_img': {
                'img_001': {1: 1, 3: 3},
                'img_002': {5: 5},
            }
        },
    }
}

ARGS = {
    'ious': ious,
    'detected_relations_per_image': detected_relations_per_image,
    'epoch_detection_results': epoch_detection_results,
    'relations_groundtruth_per_image': relations_groundtruth_per_image,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type dict — a dict output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a dict output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
get_row_overlaps Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/postprocess_table_structure.py
code served (permissive licence) · get_code("78a12b51137f2940")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family int; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# Construct row_anns: a dict mapping row numbers to annotations with 'bbox' key
# bbox format in COCO is [x0, y0, width, height]
row_anns = {
    0: {'bbox': [10, 20, 100, 50]},  # x0=10, y0=20, w=100, h=50
    1: {'bbox': [15, 100, 90, 40]},  # x0=15, y0=100, w=90, h=40
    2: {'bbox': [20, 200, 80, 30]},  # x0=20, y0=200, w=80, h=30
}

# Construct cell_ann: an annotation with 'bbox' key
# This bbox will be checked for height overlap with each row
cell_ann = {'bbox': [12, 30, 95, 45]}  # x0=12, y0=30, w=95, h=45

ARGS = {
    'row_anns': row_anns,
    'cell_ann': cell_ann,
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family int — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
is_in_region Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/stage2_structure_parser.py
code served (permissive licence) · get_code("61b73caa853c9aac")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned int, ndim 0, dtype family int, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch

torch.manual_seed(0)

# The function is_in_region takes 4 parameters:
# - bbox_x0: float, left edge of bounding box
# - bbox_x1: float, right edge of bounding box
# - left_region: tuple or list with [0] and [1] indices (left and right edges)
# - right_region: tuple or list with [0] and [1] indices (left and right edges)

# All parameters are numeric values used in comparisons and arithmetic
# No complex objects are needed

ARGS = {
    "bbox_x0": 10.0,
    "bbox_x1": 30.0,
    "left_region": [0.0, 20.0],
    "right_region": [25.0, 50.0]
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type int, ndim 0, dtype family int, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
match_text_elements_with_cell_bbox Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/postprocess_utils.py
code served (permissive licence) · get_code("63b63b787d68e2b6")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned list, ndim 1, dtype family float; no output shape is stated. The image it ran in was not recorded. If your copy returns the same type and ndim on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

# Setup code for match_text_elements_with_cell_bbox

# Define sample bounding boxes in xyformat (x0, y0, x1, y1)
cell_bbox = [10.0, 20.0, 100.0, 80.0]

# Define sample text elements with flippedbbox and content
text_elems_for_page = [
    {
        'flippedbbox': [15.0, 25.0, 95.0, 40.0],
        'content': '  Sample Text 1  '
    },
    {
        'flippedbbox': [15.0, 45.0, 95.0, 60.0],
        'content': '  Sample Text 2  '
    },
    {
        'flippedbbox': [200.0, 200.0, 250.0, 220.0],
        'content': '  Outside Text  '
    },
    {
        'flippedbbox': [20.0, 50.0, 90.0, 70.0],
        'content': '  Sample Text 3  '
    }
]

ARGS = {
    'cell_bbox': cell_bbox,
    'text_elems_for_page': text_elems_for_page
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type list, ndim 1, dtype family float — a list output has no array shape of its own; none is stated
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
prettify Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/postprocess_utils.py
code served (permissive licence) · get_code("2369b6e1b7bb0f60")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned str. The image it ran in was not recorded. If your copy returns the same type on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

from xml.etree import ElementTree

# Create a simple XML element to test the prettify function
root = ElementTree.Element("root")
child1 = ElementTree.SubElement(root, "child1")
child1.text = "Hello"
child2 = ElementTree.SubElement(root, "child2")
child2.text = "World"

ARGS = {
    "elem": root
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
none outside the standard library — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type str — a str output has no array shape of its own; none is stated — the instrument coerces any output to an array; for a str output the recorded ndim 0 and dtype family float describe that coercion, not the output, and are not stated
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
represents_int Ran
ran, violated the contract we drafted
DS3Lab/DocParser/docparser/utils/data_utils.py
pointer only (licence: MIT); the stored copy get_code serves, from aminebdj/openyolo3d, is pointer only (licence: NONE) · get_code("945af43d6054ff01")
How it ran

The body of this sample is withheld: this paper's link allows the body inline (MIT), but get_code serves the stored copy, first harvested from aminebdj/openyolo3d, and GitHub detected no licence in that repository when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded.

second_bbox_overlaps_with_first_bbox_by_percentage_xyformat Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/utils/postprocess_utils.py
code served (permissive licence) · get_code("9e75cd34e97c5ab8")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyzing the function:
# - b1: a bounding box in xyformat [x0, y0, x1, y1]
# - b2: another bounding box in xyformat [x0, y0, x1, y1]
# - percentage: a float threshold (default 0.5)
# The function checks if b2 overlaps with b1 by at least the given percentage of b2's area

# Create two bounding boxes in xyformat [x0, y0, x1, y1]
# b1 will be a larger box
b1 = [0, 0, 10, 10]  # box from (0,0) to (10,10), area = 100

# b2 will be a smaller box that overlaps significantly with b1
b2 = [5, 5, 15, 15]  # box from (5,5) to (15,15), area = 100
# overlap region: (5,5) to (10,10), area = 25
# overlap percentage of b2: 25/100 = 0.25

# Create a percentage threshold
percentage = 0.2  # 20%, so 25% overlap should return True

ARGS = {
    "b1": b1,
    "b2": b2,
    "percentage": percentage
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
value_is_in_range Ran
ran on a synthesized input
DS3Lab/DocParser/docparser/stage2_structure_parser.py
code served (permissive licence) · get_code("b38f471c4f044693")
How it ran

Ran in Syntology's sandbox on this synthesized input (drafted by a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API) and returned bool, ndim 0, dtype family float, shape () as a later run recorded it. The image it ran in was not recorded. If your copy returns the same type, ndim and shape on these inputs, it matched what we observed; nothing here checks that. It does NOT mean the output is correct: nothing has compared it to the paper.

import torch
import numpy as np

torch.manual_seed(0)

# Analyze the function:
# - value: a comparable value (number)
# - value_range: a sequence with at least 2 elements [min, max]
# The function checks if value is between value_range[0] and value_range[1]

# Create valid arguments
value = 5
value_range = [0, 10]

ARGS = {
    "value": value,
    "value_range": value_range
}
Setup block drafted by
a language model, Claude Haiku 4.5, prompt python_fixture_v1, through the batch API
Imports
numpy, torch — import names as written, not package names to install; a name the repository itself provides is not a package
Returned
type bool, ndim 0, dtype family float, shape () — shape () as a later run recorded it
dtype family
int means an integer dtype; float means any other dtype the instrument saw, bool and complex included (fixture_check.py records only these two families)
Environment
image not recorded; re-verify against syntology-harvest-sandbox:v5-deps; not recorded for this sample: the harness installs a cuda-to-cpu shim (cuda_shim.py) before every sample and no run records whether a sample needed it; seed: none by the harness; any seed is in the setup block
add_bb_into_image Not yet run
unverified (harvested, no recorded run)
DS3Lab/DocParser/docparser/objdetmetrics_lib/utils.py
code served (permissive licence) · get_code("3c195011aeae2be0")
How it ran

no invocation recorded.

getBoundingBoxes Not yet run
unverified (harvested, no recorded run)
DS3Lab/DocParser/docparser/objdetmetrics_lib/BoundingBoxes.py
code served (permissive licence) · get_code("6f20ac22161e687b")
How it ran

no invocation recorded.

getBoundingBoxesForFile Not yet run
unverified (harvested, no recorded run)
DS3Lab/DocParser/docparser/objdetmetrics_lib/BoundingBoxes.py
code served (permissive licence) · get_code("448c739d0f3a7913")
How it ran

no invocation recorded.

Also present in other papers' code

Of 17 samples lifted from this paper's linked repositories, 17 carry a function identity and 0 do not. Of the 16 distinct functions with an identity, 3 are also present in other papers whose code we hold (3 with at least one paper dated earlier, 0 the same month, 1 later) and 13 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).

Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.

FunctionAlso present inEarliest-datedOthers named
represents_int 7 other papers
1 dated earlier, 6 later
1906.01140 (2019-06)
Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds
2306.03908 (2023-06)
SAM3D: Segment Anything in 3D Scenes
2311.16054 (2023-11)
Metric Space Magnitude for Evaluating the Diversity of Latent Representations
and 4 more not named here
convertToAbsoluteValues 1 other paper
1 dated earlier
1409.1556 (2014-09)
Very Deep Convolutional Networks for Large-Scale Image Recognition
convertToRelativeValues 1 other paper
1 dated earlier
1409.1556 (2014-09)
Very Deep Convolutional Networks for Large-Scale Image Recognition

Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.

Repositories linked to this paper

Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.

Abstract

Translating renderings (e. g. PDFs, scans) into hierarchical document structures is extensively demanded in the daily routines of many real-world applications. However, a holistic, principled approach to inferring the complete hierarchical structure of documents is missing. As a remedy, we developed "DocParser": an end-to-end system for parsing the complete document structure -including all text elements, nested figures, tables, and table cell structures. Our second contribution is to provide a dataset for evaluating hierarchical document structure parsing. Our third contribution is to propose a scalable learning framework for settings where domain-specific data are scarce, which we address by a novel approach to weak supervision that significantly improves the document structure parsing performance. Our experiments confirm the effectiveness of our proposed weak supervision: Compared to the baseline without weak supervision, it improves the mean average precision for detecting document entities by 39.1 % and improves the F1 score of classifying hierarchical relations by 35.8 %.

For agents

The same record, over MCP at https://syntology.ai/mcp:

get_harvested_code_for_paper("aaai_16558")
get_code_for_paper("aaai_16558")
have("aaai_16558")

The run record, dated, one paper per request, free:

curl https://syntology.ai/api/ran/aaai_16558.json

A badge for a README (the split and the date, never a ratio):

[![Syntology run record](https://syntology.ai/api/ran/aaai_16558.svg)](https://syntology.ai/paper/aaai_16558)

Connect an agent — have() is free.